Traversability classification using super-voxel method in unstructured terrain

Soohwan Song, Sungho Jo

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

Estimating the traversability of terrain in an unstructured outdoor environment is one of the challenging issues in autonomous vehicles. When dealing with a large 3D point cloud, the computational cost of processing all of the individual points is very high. Thus voxelization methods are used extensively. In this paper, we propose a more fine-grained voxelization algorithm in the context of unstructured terrain classification. While the current shape of a voxel is a fixed-length cubic, we construct a flexible shape voxel which has spatial and geometrical properties. Furthermore, we propose a new shape histogram feature that represents the statistical characteristics of 3D points. The proposed method was tested using data obtained from unstructured outdoor environments for performance evaluation.

Original languageEnglish
Title of host publicationRobot Intelligence Technology and Applications 3 - Edition of the Selected Papers from the 3rd International Conference on Robot Intelligence Technology and Applications
EditorsWeimin Yang, Hyun Myung, Jong-Hwan Kim, Peter Sincak, Jun Jo
PublisherSpringer Verlag
Pages595-604
Number of pages10
ISBN (Print)9783319168401
DOIs
StatePublished - 2015
Event3rd International Conference on Robot Intelligence Technology and Applications, RiTA 2014 - Beijing, China
Duration: 6 Nov 20148 Nov 2014

Publication series

NameAdvances in Intelligent Systems and Computing
Volume345
ISSN (Print)2194-5357

Conference

Conference3rd International Conference on Robot Intelligence Technology and Applications, RiTA 2014
Country/TerritoryChina
CityBeijing
Period6/11/148/11/14

Keywords

  • Point cloud
  • Traversability classification
  • Unmanned vehicle
  • Unstructured terrain
  • Voxel

Fingerprint

Dive into the research topics of 'Traversability classification using super-voxel method in unstructured terrain'. Together they form a unique fingerprint.

Cite this